On 2024/10/11 8:17, Sasha Levin wrote:

This is a note to let you know that I've just added the patch titled

 Revert "mm/filemap: avoid buffered read/write race to read inconsistent data"

to the 6.6-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git%3Ba=su...

The filename of the patch is: revert-mm-filemap-avoid-buffered-read-write-race-to-.patch and it can be found in the queue-6.6 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let stable@vger.kernel.org know about it.

Hi Sasha,

The current patch is a cleanup after adding smp_load_acquire/store_release() to i_size_read/write().

In my opinion stable versions continue to use smp_rmb just fine, no need to backport the current patch to stable.

In addition, the current patch belongs to a patch set:

https://lore.kernel.org/all/20240124142857.4146716-1-libaokun1@huawei.com/

If the current patch does need to be backported to stable, then the entire patch set needs to be backported or problems will be introduced.

All the patches in the patch set are listed below:

d8f899d13d72 ("fs: make the i_size_read/write helpers be smp_load_acquire/store_release()")  4b944f8ef996 ("Revert "mm/filemap: avoid buffered read/write race to read inconsistent data"")  ad72872eb3ae ("asm-generic: remove extra type checking in acquire/release for non-SMP case")

Thanks, Baokun

commit d87bcb0965636a9b665231560ce147d06a7a18d8 Author: Baokun Li libaokun1@huawei.com Date: Wed Jan 24 22:28:56 2024 +0800

 Revert "mm/filemap: avoid buffered read/write race to read inconsistent data"
 
 [ Upstream commit 4b944f8ef99641d5af287c7d9df91d20ef5d3e88 ]
 
 This reverts commit e2c27b803bb6 ("mm/filemap: avoid buffered read/write
 race to read inconsistent data"). After making the i_size_read/write
 helpers be smp_load_acquire/store_release(), it is already guaranteed that
 changes to page contents are visible before we see increased inode size,
 so the extra smp_rmb() in filemap_read() can be removed.
 
 Signed-off-by: Baokun Li <libaokun1@huawei.com>
 Link: https://lore.kernel.org/r/20240124142857.4146716-3-libaokun1@huawei.com
 Signed-off-by: Christian Brauner <brauner@kernel.org>
 Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/mm/filemap.c b/mm/filemap.c index e6c112f3a211f..cd028f3be6026 100644 --- a/mm/filemap.c +++ b/mm/filemap.c @@ -2694,15 +2694,6 @@ ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *iter, goto put_folios; end_offset = min_t(loff_t, isize, iocb->ki_pos + iter->count);

• /*
  

•  * Pairs with a barrier in
  

•  * block_write_end()->mark_buffer_dirty() or other page
  

•  * dirtying routines like iomap_write_end() to ensure
  

•  * changes to page contents are visible before we see
  

•  * increased inode size.
  

•  */
  

• smp_rmb();
  

• /*
  • Once we start copying data, we don't want to be touching any
  • cachelines that might be contended: